1730 Diabetes Volume 68, September 2019

Targeting BCAA Catabolism to Treat Obesity-Associated Insulin Resistance

Meiyi Zhou,1 Jing Shao,1 Cheng-Yang Wu,2 Le Shu,3 Weibing Dong,1 Yunxia Liu,1 Mengping Chen,1 R. Max Wynn,2 Jiqiu Wang,4 Ji Wang,1 Wen-Jun Gui,2 Xiangbing Qi,5 Aldons J. Lusis,6 Zhaoping Li,7 Weiqing Wang,4 Guang Ning,4 Xia Yang,3 David T. Chuang,2 Yibin Wang,8 and Haipeng Sun1,8

Diabetes 2019;68:1730–1746 | https://doi.org/10.2337/db18-0927

Recent studies implicate a strong association between catabolic defect and elevated abundance of BCAAs and elevated plasma branched-chain amino acids (BCAAs) BCKAs in obesity-associated IR and provide proof-of- and insulin resistance (IR). However, a causal relationship concept evidence for the therapeutic validity of manipu- and whether interrupted BCAA homeostasis can serve as lating BCAA metabolism for treating diabetes. a therapeutic target for diabetes remain to be established experimentally. In this study, unbiased integrative pathway analyses identified a unique genetic link between obesity- The three branched-chain amino acids (BCAAs) are leucine, associated IR and BCAA catabolic expression at the isoleucine, and valine. A strong association between ele- pathway level in human and mouse populations. In genet- vated plasma BCAAs and insulin resistance (IR) has been ob/ob ically obese ( ) mice, rate-limiting branched-chain repeatedly observed in human and rodent models (1–9). a fi -keto acid (BCKA) de ciency (i.e., BCAA Moreover, longitudinal studies suggest that a high plasma and BCKA accumulation), a metabolic feature, accompa- BCAA level is predictive of the future onset of diabetes nied the systemic suppression of BCAA catabolic . – fi

METABOLISM (10 13). Circulating BCAAs are also a signi cant prognos- Restoring BCAA catabolic flux with a pharmacological tic marker associated with outcomes of diabetes interven- inhibitor of BCKA dehydrogenase (BCKDK) ( a sup- tions (4,10,14–16). On the other hand, elevated plasma pressor of BCKA dehydrogenase) reduced the abundance a of BCAA and BCKA and markedly attenuated IR in ob/ob branched-chain -keto acids (BCKAs), the products of mice. Similar outcomes were achieved by reducing protein BCAA transamination, are also associated with IR and (and thus BCAA) intake, whereas increasing BCAA intake are potentially better biomarkers for diabetes (8,9). These did the opposite; this corroborates the pathogenic roles of observations strongly suggest a causal role of disrupted BCAAs and BCKAs in IR in ob/ob mice. Like BCAAs, BCAA homeostasis in IR, which remains to be established BCKAsalsosuppressedinsulin signaling via activation experimentally. of mammalian target of rapamycin complex 1. Finally, BCAAs are essential amino acids derived from protein- the small-molecule BCKDK inhibitor significantly attenu- containing foods. The BCAA catabolic pathway, consisting ated IR in high-fat diet–induced obese mice. Collectively, of more than 40 in mitochondria, plays a pivotal these data demonstrate a pivotal causal role of a BCAA role in maintaining BCAA homeostasis in mammals (2).

1Department of Pathophysiology, Key Laboratory of Cell Differentiation and 8Departments of Anesthesiology, Medicine, and Physiology, University of Apoptosis of the Chinese Ministry of Education, Hongqiao International Institute California at Los Angeles, Los Angeles, CA of Medicine, Shanghai Tongren Hospital/Faculty of Basic Medicine, Shanghai Jiao Corresponding author: Haipeng Sun, [email protected] Tong University School of Medicine, Shanghai, China Received 26 August 2018 and accepted 29 May 2019 2Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, TX 3Department of Integrative Biology and Physiology, University of California at Los This article contains Supplementary Data online at http://diabetes Angeles, Los Angeles, CA .diabetesjournals.org/lookup/suppl/doi:10.2337/db18-0927/-/DC1. 4Department of Endocrinology and Metabolism, Ruijin Hospital, Shanghai Jiao M.Z., J.S., C.-Y.W., and L.S. contributed equally to this work. Tong University School of Medicine, Shanghai, China © 2019 by the American Diabetes Association. Readers may use this article as 5Chemistry Center, National Institute of Biological Science, Beijing, China long as the work is properly cited, the use is educational and not for profit, and the 6Departments of Medicine, Microbiology, and Human Genetics, University of work is not altered. More information is available at http://www.diabetesjournals California at Los Angeles, Los Angeles, CA .org/content/license. 7Department of Clinical Nutrition, University of California at Los Angeles, Los Angeles, CA diabetes.diabetesjournals.org Zhou and Associates 1731

BCAA catabolism is initiated by branched-chain amino- listed in Supplementary Table 1. A total of 1,690 coex- (BCAT), which facilitates a reversible pression modules were constructed from adipose, liver, transamination reaction generating BCKAs including andmuscletissuesamplesgeneratedfrommultiplehuman a-ketoisocaproic acid (from leucine), a-keto-b-methylvaleric and mouse studies (Supplementary Table 1) by using the acid (from isoleucine), and a-ketoisovaleric acid (from WGCNA package in R software (32). valine). Mitochondrial BCAT (BCAT2) is expressed ubiq- uitously and plays a main role in peripheral BCAA catab- Amino Acid Pathways and Other Canonical Pathways olism (2,17). The subsequent irreversible decarboxylation We retrieved BCAA catabolism pathways and 11 pathways of BCKAs by BCKA dehydrogenase (BCKD) is the rate- for amino acids that are not BCAAs from the Kyoto limiting step in BCAA catabolism, giving rise to CoA Encyclopedia of Genes and Genomes (KEGG) Database moieties that feed into the citric acid cycle after several (33). PPM1K and BCKDK were manually added into the reactions. In addition to substrate-dependent allosteric KEGG BCAA pathway, which was further categorized into modulation, BCKD activity is also regulated by posttrans- groups of genes specific to the degradation of leucine, lational modifications. Phosphorylation of the BCKD E1a valine, and isoleucine. For non-BCAA pathways, we also subunit by BCKD kinase (BCKDK) inhibits BCKD, whereas removed genes overlapping with BCAA. We used the amino dephosphorylation by protein phosphatase 2Cm (PP2Cm) acid pathways, along with other pathways from the KEGG activates BCKD. Loss-of-function mutations in BCAA cat- Database, to annotate all 1,690 coexpression modules; abolic genes encoding BCKD subunits and PP2Cm lead to we tested the enrichment of pathway genes using the BCKD deficiency, BCAA and BCKA accumulation, and Fisher exact test. Statistical significance was indicated maple syrup urine disease (18–21). by Bonferroni-corrected P , 0.05, fold enrichment .5, Given the critical role of the BCAA catabolic pathway in and .2 overlapping genes. maintaining BCAA homeostasis and the strong association between diabetes and elevated BCAAs and BCKAs, the Integrative Genomics Analysis Using the Mergeomics Pipeline BCAA catabolic pathway may play an important role in We used the Marker Set Enrichment Analysis (MSEA) IR pathogenesis. Individual genes (BCKDHA and PP2Cm) of the BCAA catabolic pathway have been genetically linked library in Mergeomics to determine the association of coexpression networks with human insulin traits by with IR (22–25). In this study, integrated genomic analyses leveraging human GWAS and eQTLs (34) (Fig. 1A). Spe- revealed a unique association between BCAA catabolic gene cifically, coexpression modules were first mapped to adi- expression at the pathway level and obesity-associated IR pose, liver, and muscle eQTLs in order to derive the in human and mouse populations. We then examined the corresponding representative expression SNP sets. The causal role of suppressed BCAA catabolism in IR in both P values indicating the association of the expression ob/ob and high fat–induced obese mice and the therapeutic SNPs with disease were then extracted from the filtered potential of targeting BCAA catabolism to treat type 2 diabetes. summary-level statistics as described above. We assessed the significance of enrichment for SNPs indicating mod- RESEARCH DESIGN AND METHODS erate to strong risk for a trait within each module using a x2 like statistic followed by multiple-testing correction in -Wide Association Studies for Insulin- Related Traits order to estimate false discovery rate (FDR) (34). Modules with an FDR ,5% and P , 0.05 were considered to be Publicly available human genome-wide association study significantly or suggestively associated with the respective (GWAS) data sets for fasting glucose, fasting insulin, trait. and IR (unadjusted and adjusted for BMI) were retrieved from large meta-analysis consortia including MAGIC (the To test whether BCAA genes play key regulatory roles in the trait-associated BCAA modules, we used the weighted Meta-Analyses of Glucose and Insulin-related Traits Con- Key Driver Analysis library in Mergeomics to identify key sortium) (26) and the GENESIS (Genetics of Insulin Sen- drivers of the significant modules. Key drivers of a given sitivity) Consortium (27). After retrieving summary-level module were defined as genes whose neighboring sub- statistics for all single nucleotide polymorphisms (SNPs), network exhibited significant enrichment (FDR ,1%) for we removed SNPs with a weak association (,80%). The 2 member genes in that module. Adipose Bayesian networks remaining SNPs with high linkage disequilibrium (r . 0.5) used in key driver analysis were constructed through the were filtered by using a previously described method (28). Linkage disequilibrium data for European ancestry was use of a previously developed method (35,36). obtained from HapMap3 (29) and the 1000 Genomes Project (30). Comprehensive lists of human cis–expression Gene-Trait Correlation in the Hybrid Mouse Diversity quantitative trait loci (eQTL) from human adipose, liver, Panel and muscle tissues were accessible from our Mergeomics Pearson correlations of the expression of BCAA genes web server (31). Cis-eQTL were defined as eQTL in which in liver, adipose, and muscle tissues with fasting glu- the associated SNP and gene pairs are within 1 MB of each cose, fasting insulin, and HOMA-IR measurements other. Details of the eQTL data sets used in the study are were retrieved from the Hybrid Mouse Diversity Panel 1732 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

Figure 1—Integrative genomic analyses associate the BCAA catabolic pathway with IR-related traits in human populations. A: The integrative genomics workflow we used to investigate the association of BCAAs with IR-related traits in humans. Specifically, human GWAS were integrated with eQTLs and coexpression networks matched by tissue, and then analyzed using the Mergeomics pipeline in order to identify coexpression modules that showed significant genetic association with IR-related clinical traits. Coexpression modules with significant over-representation of BCAA genes among the module genes were then annotated as BCAA modules. B: Number of BCAA modules with significant trait association (FDR ,5% or P , 0.05, as assessed by using MSEA in the Mergeomics pipeline). Numbers in the bars indicate the fold enrichment of BCAA modules among all significant coexpression modules for the corresponding trait. Statistical significance of enrichment of BCAA modules among all significant modules was determined by using the Fisher exact test. Details of enrichment test are in Supplementary Table 2. C: Comparison of tissue origin distribution of BCAA modules and all coexpression modules significantly associated with fasting insulin and IR (BMI unadjusted) at FDR ,5% and P , 0.05. Significance of differences in the mean correlation strength between gene categories was calculated by using the Student t test. *P , 0.05; **P , 0.01; ***P , 0.001; ****P , 0.0001.

(HMDP) (https://systems.genetics.ucla.edu/data/hmdp), Animals Co. Ltd, Shanghai,China.AnimalsfromThe which comprises ;100 strains of mice fed a high-fat diet Jackson Laboratory were used for our investigation of 3,6- (HFD) (37,38). We assessed clinical traits using 8–12 mice per dichlorobenzo[b]thiophene-2-carboxylic acid (BT2), and ani- strain, whereas we profiled gene expression using 3 mice per mals from SLAC were used for other studies. All mice were strain per tissue. housed at 22°C under a 12-h light/12-h dark cycle, with free access to water and standard chow. An isocaloric normal Animals protein diet (NPD) (20% protein; TD.91352) and a low- Male ob/ob mice or male wild-type C57BL/6 mice were protein diet (LPD) (6% protein; TD.90016) (Supplementary purchased from The Jackson Laboratory or SLAC Laboratory Table 5) were purchased from Envigo Teklad Diet. A normal diabetes.diabetesjournals.org Zhou and Associates 1733

BCAA diet (A11072001) and a low-BCAA diet (60% less measure BCAA concentrations. To determine BCKA con- BCAAs than the normal BCAA diet; A12030802) (Supplemen- centrations, plasma, a solution of [13C] a-ketoisovaleric tary Table 6) were purchased from OpenSource Diets. Tissue acid, and o-phenylenediamine were mixed and incubated samples from white adipose tissue (epididymal fat), skeletal at 80°C for 20 min. Na2SO4 and ethyl acetate were added muscle (soleus/gastrocnemius), and liver were quickly har- to the cooled solutions, which were then centrifuged. The vested and frozen in liquid nitrogen and maintained at 280°C upper organic phase was re-extracted, pooled with the until processed. All animal procedures were carried out in first, and dried under mild heat (40°C). The samples were accordance with the guidelines and protocols approved by the resuspended in MeOH and 5 mmol/L NH4 acetate (50:50) Committee for Humane Treatment of Animals at Shanghai and analyzed by liquid chromatography–tandem mass Jiao Tong University School of Medicine, the University of spectrometry. Texas Southwestern Medical Center Institutional Animal Care and Use Committee, or the University of California at Los Glucose and Insulin Tolerance Tests Angeles Institutional Animal Care and Use Committee. Mice were deprived of food for 6 h. For the insulin tolerance test, mice were injected intraperitoneally with insulin (0.75 Western Blotting Analysis units/kg body weight; Sigma). For the glucose tolerance test, Antisera against the BCKD E1a subunit were a gift from mice were injected intraperitoneally with D-glucose (1.5 g/kg Dr. Yoshiharu Shimomura (Nagoya Institute of Technology, body weight; Sigma). Blood glucose concentrations were Nagoya, Japan). We purchased other antibodies against measured in tail blood by using a portable glucometer pBCKD E1a (Novus Biologicals); BCKDE2 (Thermo Fisher (Johnson & Johnson) at the times indicated after Scientific); and phospho-Akt (Thr308), Akt, phospho- injection in figures. Plasma insulin was measured with p70S6K (Thr389), total p70S6K, tubulin, GAPDH, and MILLIPLEX Multiplex Immunoassay Kits (MMHMAG- b-actin (Cell Signaling Technology). Densitometry was 44K-14; Merck Millipore, Darmstadt, Germany). To allow performed with ImageJ software. us to examine in vivo insulin signaling, mice were deprived of food for 6 h and injected intraperitoneally with insulin RNA Isolation and Quantitative RT-PCR (2 units/kg body weight) for 10 min. Total RNA was extracted from the various tissues by using the TRIzol Reagent (Invitrogen), according to the manu- Mouse Treatment With the BCKDK Inhibitor BT2 facturer’s instructions. Total RNA (2 mg) was reverse The animals were fed an HFD (60% kcal from fat) (catalog transcribed by using random primers and Maloney murine no. D12492; Research Diets) for ;12 weeks starting at the leukemia virus (Promega). Each cDNA sample was ana- age of ;6 weeks in order to induce obesity in wild-type lyzed in triplicate with the Applied Biosystems Prism mice (diet-induced obesity [DIO]). For ob/ob mice, the 7900HT Real-Time PCR System using Absolute SYBR treatment started at 8 weeks of age. Obese mice (DIO Green (Applied Biosystems). The primer sequences are mice or ob/ob mice) were randomized into two groups and listed in Supplementary Table 4. received either the vehicle or BT2 treatment. BT2 was dissolved in DMSO and diluted in 5% DMSO, 10% Cre- Metabolomic Analysis mophor EL, and 85% 0.1 mol/L sodium bicarbonate (pH The global metabolomic analysis was carried out by Metab- 9.0) for delivery. Mice received via oral gavage daily doses olon, Inc. (Durham, NC), as described previously (39). The of BT2 (40 mg/kg/day) or an equal volume of vehicle for Welch two-sample t test was used to test whether two 8–10 weeks. BCKD activity was analyzed as described unknown means were different from two independent previously (40). populations. The wild-type mice were fed the NPD (20% protein by weight) and ob/ob mice were fed the NPD or Statistics LPD (6% protein by weight) for 4 weeks beginning at Unless otherwise specified, statistical analyses were per- 10 weeks of age. For the BCAA group (LPD + BCAA), formed with the two-sided Student t test (two groups) and drinking water was supplemented with BCAA (3 mg/mL) one-way ANOVA (more than two groups) or two-way beginning after 2 weeks of eating the LPD; this supple- ANOVA (tolerance tests), followed by a Tukey post hoc mentation lasted 2 weeks. test, where appropriate, using GraphPad Prism software. Data were calculated as the mean 6 SD unless otherwise indicated. A P value ,0.05 was considered to be statisti- Determination of BCAA and BCKA Concentrations With cally significant. Mass Spectrometry Plasma was precleared of protein by using ethanol. The supernatant was lyophilized and resuspended. Standards RESULTS and samples were diluted in butanolic HCL, heated to 60°C BCAA Catabolic Pathway Is Genetically Associated for 20 min, and then dried in a speed vacuum. The samples With Obesity-Associated IR in Human Populations were resuspended in distilled H2O and acetonitrile (50:50) To explore the connection between specific metabolic containing 0.1% formic acid, and analyzed by liquid pathways and IR in an unbiased manner, we leveraged chromatography–tandem mass spectrometry in order to existing genetics and functional genomics data sets from 1734 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019 human and mouse populations using an integrative frame- Consistent with the findings from the aforementioned work (Fig. 1A). human studies, the expression levels of BCAA metabolic For the human-focused analysis, we mainly used GWAS genes in the adipose tissue exhibited a strong negative of IR from the GENESIS Consortium (27) and glycemic correlation with fasting insulin level and IR (HOMA-IR), traits from MAGIC (26) to capture genetic association but not fasting glucose level, in the HMDP cohort (Fig. 2A signals. Instead of focusing on individual genetic signals and B). The expression levels of a smaller set of hepatic in individual genes, we focused on the aggregate behavior BCAA catabolic genes showed either a positive or a negative of groups of functionally related genes by integrating the association with fasting glucose, fasting insulin, and IR GWAS data sets with tissue-specific eQTLs and coexpres- traits (Fig. 2B). The strength of the correlation in liver was sion network modules containing sets of coregulated genes weaker than that in the adipose tissue (Fig. 2B) and was in IR-relevant tissues—including adipose tissue, liver, and not significantly different from the genes in amino acid muscle (Fig. 1A)—using the MSEA library in the Mergeo- (non-BCAA) pathways (Fig. 2A). The muscle tissue, on the mics pipeline (34). Among all significant coexpression other hand, did not exhibit specific correlations between modules for fasting insulin and IR traits at either an BCAA catabolic gene expression and glycemic traits (Fig. FDR ,5% (correcting for multiple testing) or P , 0.05 2A and B). These data from the HMDP mouse cohort study (uncorrected), we observed significant over-representation further support the association between IR and BCAA of modules annotated with the BCAA catabolic pathway, catabolic gene expression observed in the human popula- among other well-established IR-related processes such as tion, substantiating a common genetic feature related to glucose metabolism, oxidative phosphorylation, and in- the BCAA catabolic pathway in diabetes. flammation (Fig. 1B and Supplementary Table 2). For instance, 57 of a total of 1,690 modules were found to BCKD Deficiency and BCAA/BCKA Accumulation be significantly associated with IR. Among these, seven Characterize BCAA Catabolic Defect in Obese ob/ob modules were enriched for BCAA catabolic genes, repre- Mice senting a 4.2-fold enrichment for BCAA-related modules We next investigated the relationship between the BCAA (enrichment P = 1.8e24) (Fig. 1B and Supplementary catabolic pathway and obesity-associated IR in a well- Table 2). In contrast, the BCAA catabolic pathway was established experimental obese/diabetic animal model, not among coexpression modules associated with fasting leptin-deficient (ob/ob) mice (Supplementary Fig. 3). Ex- glucose or BMI-adjusted IR, suggesting an adiposity- amination of gene expression revealed systemic down- dependent association between the BCAA catabolic path- regulation of the BCAA catabolic pathway in white way and IR (Fig. 1B). No association was observed for adipose tissue and liver, but not in skeletal muscle, in metabolic pathways for amino acids that are not BCAAs ob/ob mice (Fig. 3A). These expression patterns corre- (Fig. 1B). Details of the BCAA catabolic genes, correspond- sponded to the genetic association between IR and the ing eQTLs, and P values for trait associations for the BCAA catabolic pathway (particularly the negative corre- modules are provided in Supplementary Table 3. lation in adipose tissue) identified by the genomic analyses After matching the tissue origin of coexpression mod- in human and mouse populations (Figs. 1 and 2). ules and eQTLs, we found that the BCAA modules asso- We performed a targeted metabolomics analysis to ciated with IR were primarily from the adipose tissue, not characterize BCAA catabolism in ob/ob mice by examining skeletal muscle or liver (Fig. 1C). In addition, we used the the abundance of BCAAs and their metabolites in plasma weighted Key Driver Analysis library in the Mergeomics and various tissues (Fig. 3B). Fasting plasma levels of pipeline to pinpoint potential key drivers (28,41–43), BCAAs were unexpectedly not higher in obese mice (Fig. revealing BCAA catabolic genes as key drivers for all 3C). Mass spectrometry also detected similar specific BCAA modules. These results support BCAA metabolic measurements of plasma BCAA levels in food-deprived genes as playing a pivotal regulatory role in the wild-type and ob/ob mice (Supplementary Fig. 4). BCAA IR-associated BCAA modules (Supplementary Fig. 1). To- abundances were, however, higher in white adipose tissue, gether, these unbiased genomics analyses from human lower in liver, and unchanged in skeletal muscle in food- populations reveal a unique and strong association be- deprived obese mice relative to lean mice (Fig. 3C). The tween obesity-associated IR and the BCAA catabolic fasting plasma BCKA concentrations were significantly pathway. higher, in accordance with previous reports (8,9). Signif- icant elevation of BCKAs in white adipose tissue and BCAA Catabolic Gene Expression at the Pathway Level trends of elevation in liver and skeletal muscle were Is Correlated With IR in the HMDP also detected in food-deprived ob/ob mice (Fig. 3C). On To connect the human genomic analysis results to mouse the other hand, the concentrations of BCKD metabolites models in which IR is typically studied, we evaluated the were significantly lower in plasma, liver, white adipose association of the expression levels of BCAA catabolic tissue, and probably skeletal muscle of food-deprived ob/ob genes in adipose, liver, and muscle with IR-relevant traits mice (Fig. 3C). in ;100 genetically divergent strains of mice fed with The elevation of plasma BCAA levels has been viewed as an HFD in the HMDP (37,38) (Supplementary Fig. 2). a metabolic hallmark of IR in obesity (2,4–9). The absolute diabetes.diabetesjournals.org Zhou and Associates 1735

Figure 2—The BCAA catabolic pathway shows a strong correlation with IR in a mouse population. A: Comparison of correlation strengths of the expression of BCAA pathway genes, non-BCAA amino acid pathway genes, and all genes with IR-related traits in mice. The correlation data between tissue-specific expression profiling and measurements of clinical traits were extracted from the HMDP, which contains ;100 strains of genetically divergent mice fed an HFD. The mean and SE of the absolute values of Pearson correlations between each gene within a gene category (BCAA, non-BCAA, all genes) and a trait are shown. The significance of differences in the mean correlation strength between gene categories was calculated by using the Student t test. *P , 0.05; **P , 0.01; ***P , 0.001; ****P , 0.0001. B: Correlation of individual BCAA catabolic genes in different tissues for fasting glucose, fasting insulin, and HOMA-IR in HMDP mice fed an HFD. Red indicates a positive correlation, whereas blue indicates a negative correlation. *P , 0.05 and **P , 0.05 after Bonferroni correction for the number of genes and traits. 1736 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

Figure 3—Systematic downregulation of the BCAA catabolic pathway leads to a BCAA catabolic defect in ob/ob mice. A: Quantitative PCR results of BCAA catabolic genes in white adipose tissue, skeletal muscle, and liver in lean wild-type (WT) mice (n = 4) and ob/ob mice (n =4) deprived of food for 6 h. B: Illustration of the partial BCAA catabolic process with enzymes, intermediates, and derivatives. C: Relative levels of BCAAs and their metabolites in plasma and tissues of lean WT mice (n =6–8) and ob/ob mice (n = 8). Male mice, age 14 weeks, were deprived of food for 6 h before being sacrificed. D: Serum concentrations of BCAAs and BCKAs in lean WT mice (n = 8) and ob/ob mice (n = 10). Male mice, age 8 weeks, were deprived of food from 8:00 A.M. to 5:00 P.M. and then supplied with chow diet for 1 hour. Blood (100 mL) was collected at 6:00 P.M. from the orbital sinus by using capillary tubes for serum collection. *P , 0.05 and **P , 0.01 vs. WT mice. ΚΙC, a-ketoisocaproic acid; KIV, a-ketoisovaleric acid; KMV, a-keto-b-methylvaleric acid. diabetes.diabetesjournals.org Zhou and Associates 1737 increases are usually moderate, however, and they become 1 (mTORC1) activity, and the catabolic flux may affect smaller or even disappear under fasting conditions glucose and lipid oxidation, both of which have been (6,44,45). Our data show that, in fed animals, plasma implicated as possible contributors to IR (2,50). We BCAA levels are significantly higher in ob/ob obese mice showed that BT2 treatment reduced BCAA levels while than in lean mice (Fig. 3D), and the plasma BCKA levels enhancing catabolic flux. To assess further which of these were even higher in ob/ob obese mice (Fig. 3C and D). The two changes contributes to the improved insulin sensitiv- magnitudes of BCKA elevations were more pronounced ity in ob/ob mice, we examined the influences of various than those of BCAAs in different settings, reflecting BCKD dietary manipulations that alter BCAA abundance and the deficiency. catabolic flux in different ways. We used an isocaloric LPD Therefore, although BCAA catabolic genes are sup- (6% protein by weight), instead of an NPD (20% protein by pressed at the pathway level in various tissues, systemic weight), to reduce BCAA intake and consequently the load BCAA catabolism in obese ob/ob mice is characterized by to the catabolic flux. The LPD significantly reduced the adeficiency of the rate-limiting BCKD, accompa- plasma levels of BCAAs and BCKAs (Fig. 5A) and markedly nied by reduced catabolic flux and by BCAA/BCKA accu- improved systemic glucose tolerance (Fig. 5B). In contrast, mulation. Diminished BCKD protein expression and increasing dietary intake of BCAAs by administering them increased BCKD E1a phosphorylation further support in water increased plasma levels of BCAAs/BCKAs (Fig. 5A) BCKD deficiency in obese ob/ob mice (Supplementary and significantly exacerbated the systemic glucose toler- Fig. 5), in agreement with previous studies (8,9,45–49). ance of ob/ob mice being fed the LPD (Fig. 5B). The insulin tolerance test showed that the LPD dramatically improved A BCAA Catabolic Defect Contributes to the systemic insulin sensitivity in ob/ob mice, whereas BCAA Pathogenesis of IR in ob/ob Mice supplementation significantly impaired insulin sensitivity a BCKDK phosphorylates the E1 subunit of BCKD and in mice fed the LPD (Fig. 5C). The hyperinsulinemia in fi inhibits its activity (2). BT2 is a speci c and potent in- ob/ob mice was significantly attenuated by the LPD but hibitor of BCKDK (39,40). We next assessed whether the aggravated by BCAA supplementation (Fig. 5D). Fasting BCAA catabolic defect played a causal role in obesity- plasma glucose levels were not affected by the LPD or associated IR by using BT2 to restore BCAA catabolism BCAA supplementation (Fig. 5B). The body weight of ob/ob fi in obese mice. BT2 administration signi cantly enhanced mice was slightly reduced by the LPD, whereas BCAA ob/ob BCKD activities in various tissues of mice, including supplementation did not affect food intake or body weight A liver, skeletal muscle, and white adipose tissue (Fig. 4 ). (Fig. 5E and Supplementary Fig. 9). The LPD significantly fi BT2 treatment signi cantly reduced plasma BCAA levels in reduced plasma levels of multiple amino acids; BCAAs were ob/ob the obese mice, but the magnitudes of reductions in the only ones reversed by BCAA supplementation (Sup- plasma BCKA concentrations were far more pronounced plementary Fig. 10). These results show that the BT2 B (Fig. 4 ). Importantly, BT2-treated mice displayed signif- treatment and reducing BCAA intake both attenuate IR icantly improved glucose tolerance and insulin sensitivity through a smaller abundance of BCAAs and BCKAs. On the C D (Fig. 4 and ), accompanied by the attenuation of hyper- other hand, BT2 treatment enhanced the BCAA catabolic E insulinemia (Fig. 4 ). BT2 treatment reduced the phos- flux, whereas reducing BCAA intake probably decreased a phorylation of BCKD E1 in liver and induced the the flux. In addition, increasing BCAA intake promoted IR a expression of BCKD E1 and BCKDK protein in skeletal with more BCAAs/BCKAs and higher catabolic flux. Col- muscle (Supplementary Fig. 6). BT2 also increased the lectively, all these data are consistent with the notion that expression of numerous distal genes in the BCAA catabolic more BCAAs/BCKAs, rather than the status of catabolic pathway in skeletal muscle and adipose tissue (Supple- flux, plays a key role in the onset of IR in obese ob/ob mice. mentary Fig. 7). Among the 17 amino acids that are not BCAAs, tryptophan was the only one to show a reduced Both BCAAs and BCKAs Can Impair Insulin Signaling concentration in plasma in the BT2-treated mice (Supple- We next examined insulin signaling in food-deprived mice mentary Fig. 8); this reduction is potentially due to in- in the aforementioned experimental settings. Compared creased tryptophan transport into tissues when BCAA with the response (indicated by the fold induction of levels are reduced (21). BT2 treatment did not affect pAKT-Thr308) in ob/ob mice fed the NPD, insulin signaling ob/ob F body weight or food intake in the mice (Fig. 4 in ob/ob mice fed the LPD was significantly higher in G and ). These results strongly support a causal role for the skeletal muscle (7.6 vs. 1.34 fold), liver (2.55 vs. 1.31 ob/ob BCAA catabolic defect in the onset of IR in obese fold), and white adipose tissue (2.61 vs. 1.37 fold), in- mice. dicating enhanced insulin sensitivity (Fig. 6A). In contrast, BCAA supplementation impaired insulin response in liver, Elevated BCAAs and BCKAs Are Pathogenic Factors white adipose tissue, and, to a much lesser extent, skeletal for IR in ob/ob Mice muscle in ob/ob mice fed the LPD (Fig. 6B). Similar to the The BCAA catabolic defect results in BCAA/BCKA accu- LPD, BT2 treatment significantly enhanced AKT signaling mulation and impaired BCAA catabolic flux. BCAA can in response to insulin, indicated by the higher fold in- enhance mammalian target of rapamycin complex duction of pAKT-Thr308 in various tissues (Fig. 6C). The 1738 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

Figure 4—The BCKDK inhibitor restores BCAA catabolism and attenuates IR in ob/ob mice. ob/ob mice were treated with the vehicle (Ctrl) or BT2 (at 10 weeks [A, B, and E]or4–6 weeks [C and D]) by oral gavage (n = 6 mice in each group). We analyzed BCKD activity in various tissues (A), fasting plasma levels of BCAAs (B, left) and BCKAs (B, right), glucose tolerance test results (C), insulin tolerance test results (D), fasting plasma insulin level (E), body weight (F), and food intake (G). *P , 0.05; **P , 0.01; ***P , 0.005; ****P , 0.0001 vs. Ctrl. ΚΙC, a-ketoisocaproic acid; KIV, a-ketoisovaleric acid; KMV, a-keto-b-methylvaleric acid; WAT, white adipose tissue. diabetes.diabetesjournals.org Zhou and Associates 1739

BCAA-enhanced activation of mTORC1 has been impli- cated as an important mechanism underlying the impairment of insulin signaling (1). The LPD and BT2 treatment reduced the basal or insulin-stimulated mTORC1 activity (as indicated by p70S6K phosphorylation on Thr389), or both, in all three tissues, consistent with an enhanced response to insulin (Fig. 6A and C) and reduced BCAA levels (Figs. 4B and 5A). On the other hand, BCAA sup- plementation increased the amount of BCAAs in ob/ob mice fed the LPD (Fig. 5A) and enhanced the insulin- stimulated mTORC1 activity in skeletal muscle and probably adipose tissue (Fig. 6B). Of note, BCAA supplementation impaired insulin signaling without enhancing mTORC1 activity in liver in ob/ob mice fed the LPD. As the abundances of BCKAs and BCAAs changed con- cordantly (Figs. 4B and 5A), we also examined the direct influence of BCKAs on insulin signaling in cultured cells. Interestingly, similar to BCAAs, BCKAs impaired insulin signaling while increasing insulin-stimulated mTORC1 ac- tivity (Fig. 6D). Further examination showed that BCKAs suppressed basal AKT phosphorylation while augmenting mTORC1 activity in the absence of BCAAs in culture medium (Fig. 6E). The mTORC1 inhibitor rapamycin abol- ished BCKA-repressed AKT phosphorylation, indicating an mTORC1-dependent action (Fig. 6F). These data suggest that BCKAs, in addition to BCAAs, may also contribute to impaired insulin signaling in obese ob/ob mice.

The Pharmacological Inhibitor of BCKDK Attenuates IR in DIO Mice Given the importance of the BCAA catabolic defect in the pathogenesis of obesity-associated IR, enhancing BCKD activity through small-molecule treatment may represent a new pharmacological approach to treating diabetes. We assessed the effect of the BCKDK inhibitor BT2 on insulin sensitivity in wild-type mice with DIO, a model frequently used to mimic the metabolic impairments of obesity in humans. As expected, BT2 treatment augmented BCKD activities in various tissues of DIO mice and reduced — ob/ob ob/ob Figure 5 Altering BCAA intake affects IR in mice. mice plasma BCAA and BCKA levels (Fig. 7A and B). More were fed an NPD (20% protein by weight) or an LPD (6% protein by weight) for 4 weeks beginning at 10 weeks of age. For the BCAA importantly, BT2 treatment improved glucose tolerance, group (LPD + BCAA), supplementation of BCAAs in drinking water insulin sensitivity, and hyperinsulinemia in DIO mice (Fig. (3 mg/mL) was started after mice had eaten the LPD for 2 weeks and 7C–E) without affecting body weight (Fig. 7F). These lasted 2 weeks. We analyzed fasting plasma levels of BCAAs and results demonstrate that restoring BCAA catabolic activity metabolites (n = 8 mice; *P , 0.05 vs. NPD; &P , 0.05 vs. LPD) (A), glucose tolerance test results (n 5 6–13 mice; **P , 0.01, LPD vs. with a small-molecule BCKDK inhibitor is effective to NPD; &P , 0.05, LPD vs. LPD 1 BCAA) (B), insulin tolerance test improve insulin sensitivity in DIO mice. This in vivo results (n 5 6 mice in each group; *P , 0.05, ***P , 0.001, LPD vs. proof-of-concept demonstration validates targeting the NPD; &P , 0.05, LPD vs. LPD 1 BCAA) (C), fasting plasma insulin levels (n =8–12 mice; *P , 0.05; &P , 0.05) (D), and body weight (n = BCAA catabolic defect as a potential therapeutic approach 14 mice in each group; **P , 0.01). to attenuating obesity-associated IR and diabetes.

DISCUSSION This study examines the causal role of disrupted BCAA homeostasis in obesity-associated IR and the therapeutic enhanced response to insulin in the LPD and BT2 groups potential of targeting the elevated amounts of BCAAs and was primarily attributable to the lower basal AKT activity, BCKAs. The unbiased integrative genomic analyses reveal consistent with the lower fasting insulin levels in the same a specific and robust association between the BCAA cat- groups (Figs. 4E and 5D). abolic pathway and obesity-related IR in both human and 1740 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

Figure 6—BCAAs and BCKAs contribute to impaired insulin signaling in ob/ob mice. A–C: Representative immunoblots for specific proteins, created by using tissue lysates from skeletal muscle, liver, and white adipose tissue of mice without or with insulin (Ins) injection for 10 min following 6 h of food deprivation. ob/ob mice were fed an NPD (20% protein by weight) or an LPD (6% protein by weight) for 4 weeks beginning at 10 weeks of age (A). BCAA supplementation (LPD + BCAA or LPD/amino acids) in drinking water (3 mg/mL) was started after mice had consumed the LPD for 2 weeks and lasted 2 weeks (B). ob/ob mice were treated with the vehicle (Ctrl) or BT2 by oral gavage for 5 weeks (C). The graphs below the immunoblots in A–C present densitometric values of the bands. D–F: Representative immunoblots for specific proteins, created by using cell lysates. 3T3-L1 cells were treated with FBS- and BCAA-free DMEM for 1–2hbeforeBCAA(500 mmol/L), BCKA (500 mmol/L), or rapamycin (100 nmol/L) treatment for 1 h (D and F) or various times (E), followed by insulin treatment (10 nmol/L) for 1 h (D). *P , 0.05; **P , 0.01; ***P , 0.005; ****P , 0.0001. diabetes.diabetesjournals.org Zhou and Associates 1741

Figure 7—The BCKDK inhibitor enhances BCAA catabolism and attenuates IR in DIO mice. DIO mice were treated with the vehicle (Ctrl) or BT2 by oral gavage for 8 weeks (n = 7 mice in each group). We analyzed BCKD activity in various tissues (A), fasting plasma levels of BCAAs (B, left) and BCKAs (B, right), glucose tolerance test results (C), insulin tolerance test results (D), fasting plasma insulin level (E), and body weight (F). *P , 0.05, **P , 0.01, ***P , 0.005, ****P , 0.0001, all vs. Ctrl. ΚΙC, a-ketoisocaproic acid; KIV, a-ketoisovaleric acid; KMV, a-keto- b-methylvaleric acid.

mouse populations. In genetically obese mice, characteristic supplementation increases BCAA/BCKA abundances and BCKD deficiency and BCAA/BCKA buildup accompany the promotes IR in ob/ob mice. These results together demon- suppression of BCAA catabolic genes at the pathway level. strate the pathogenic role of elevated BCAAs/BCKAs in IR in Correcting this BCAA catabolic defect abolishes BCAA/ obese mice. Similar to BCAAs, BCKAs also can impair insulin BCKA elevation and attenuates IR in obese ob/ob mice, signaling. Finally, directly targeting the BCAA catabolic supporting a causal role for the BCAA catabolic defect in IR defect with a small-molecule BCKDK inhibitor reduces onset in obese mice. Furthermore, reducing protein (and BCAA/BCKA abundances and attenuates IR in DIO mice. thus BCAA) intake effectively reduces BCAA/BCKA abun- These findings suggest that the BCAA catabolic defect and dances and improves insulin sensitivity, whereas BCAA elevated BCAAs/BCKAs play causal roles in the onset of 1742 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019 obesity-associated IR and provide a valid target for dietary remodeling in adipose tissue. As adipose tissue plays or pharmacological interventions to treat diabetes. a critical role in systemic lipid and glycemic metabolism, Several studies have established genetic links between changes in adipose physiology may in turn contribute to individual BCAA catabolic genes (BCKHA or PP2Cm) and local and systemic IR in obese mice. It is also possible, dysfunctional glucose metabolism (9,22–25). Instead of however, that an adipose BCAA catabolic defect functions individual genes, the comprehensive genomic analyses in as a main contributor to the systemic BCAA/BCKA accu- this study identified a genetic link between IR and the mulation that in turn promotes IR in various tissues. In entire BCAA catabolic pathway. The significant and unique either case, BT2 can enhance BCAA catabolism in heart, genetic association is based on information about both liver, kidney, and other tissues, relieving the systemic and genetics and gene expression. The latter is of particular adipose BCAA catabolic defect, irrespective of the site, and interest because nutritional status in obesity profoundly thus improving systemic insulin sensitivity; thus, BT2 affects BCAA catabolic gene expression. Numerous studies should be considered for therapy. have shown aberrant BCAA catabolic gene expression in Hepatic BCAA metabolism demonstrates weak and in- animals and humans that are obese and have diabetes consistent correlations with IR. Directional discrepancies (4,8,9,45–49). Interestingly, a study of monozygotic twins of correlations are observed between liver BCAA catabolic discordant for BMI, which aimed to profile obesity-altered genes and fasting insulin level and IR traits. Some BCAA biological pathways in identical genetic backgrounds, catabolic genes in liver are positively correlated with revealed significant downregulation of the BCAA adipose fasting glucose levels. The molecular details and functional catabolic pathway in obese cotwins, which correlated implications for this discrepancy are unclear. We speculate closely with IR and hyperinsulinemia (46). Thus, BCAA that some positive correlations between BCAA genes in catabolism can be impaired by obesity status and linked liver and glycemic traits in the HMDP study are likely with IR. In the bioinformatics study, we performed both a reactive rather than upstream causal process in the BMI-adjusted and BMI-unadjusted analyses to look for development of hyperglycemia or IR. Indeed, it is not IR-associated pathways that are independent of BMI uncommon to observe directional discrepancies in gene- (based on BMI-adjusted analysis) and that can be depen- trait correlations in various tissues and experimental set- dent on or independent of BMI (based on unadjusted tings because of feedback regulations and systems-level analysis). That the genetic association between the interactions between tissues. Meanwhile, the key enzymes BCAA catabolic pathway and IR is mainly observed in determining the BCAA metabolic pattern are either neg- the unadjusted analysis and disappeared in the BMI- atively (DBT) or not significantly (BCAT2, BCKHDA, adjusted analysis strongly supports an obesity-dependent BCKDHB, BCKDK) associated with glycemic traits. In relationship. Our study, together with other data, indicates addition, our human genetics analysis does not support that BCAA catabolism is suppressed in obesity, which con- an enrichment of genetic signals in liver BCAA modules for tributes to IR. any of the glycemic traits examined. Therefore, although it The special link between the adipose BCAA catabolic is possible that the BCAA catabolic pathway in liver also pathway and IR is intriguing. Studies in humans and plays a role in glycemic regulation, the genetic evidence is rodents have indicated that an impaired adipose BCAA not as strong as that for a role in adipose tissues and catabolic pathway contributes to the rise of plasma BCAAs/ requires further investigation. BCKAs and is associated with IR in obesity (4,8,9,45–49). A high level of BCAAs in blood has been viewed as In this study, the genetic association between the BCAA a metabolic hallmark of IR and diabetes in numerous catabolic pathway and obesity-associated IR is mainly studies (2,4–9). Thus, it was unexpected that the fasting observed when gene expression in adipose tissue is con- plasma BCAA levels were not elevated in ob/ob mice. sidered in the genomics analyses. In ob/ob mice, adipose Plasma BCAA abundances were higher in fed obese tissue shows the most dramatic loss of BCAA catabolic ob/ob mice, suggesting that they are strongly affected by genes and the most significant accumulation of intratissue both the dietary intake and degradation in tissues. Sim- BCAAs and BCKAs, when compared with the liver and ilarly, it has been shown that BCAA levels are significantly skeletal muscle. Therefore, BCAA catabolic activity in increased in fed obese mice, although the differences adipose tissue seems to play a unique role in regulating dramatically decline or even disappear when the mice systemic BCAA homeostasis and IR in obesity. However, it are deprived of food (45). On the other hand, the eleva- remains unclear how adipose tissue contributes signifi- tions of plasma BCKAs are significant in both food- cantly to systemic BCAA catabolism despite its relatively deprived and fed obese ob/ob mice, in agreement with low BCKD activity. It is possible that BCAA catabolic previous studies indicating higher plasma BCKA as a better capacity of tissue is determined not only by BCKD activity biomarker for diabetes than BCAA (8,9). In addition, but also by other factors such as BCAA/BCKA uptake from BCKAs are catabolites generated by tissues. Elevated blood by tissues. On the other hand, as a main contributor BCKA levels in blood reflect a BCKD deficiency in tissues to obesity-associated IR, an adipose BCAA catabolic defect of obese mice. may exert its effect locally or systemically. The local BCAAs can trigger different and even opposite meta- accumulation of BCAA/BCKA may trigger functional bolic effects in mammals, depending on their catabolic and diabetes.diabetesjournals.org Zhou and Associates 1743 anabolic states (51). Thus, a causal relationship between BCAA oxidation from the liver and fat toward muscle (55). disrupted BCAA metabolism and obesity-associated IR can In this study, BT2 reduced BCAA/BCKA abundances and be addressed only in animals that are already obese. Pre- improved insulin sensitivity while enhancing BCAA cata- vious studies examining the effects of BCAA intake on bolic activities in various tissues including liver and skele- metabolism in obese animals have fallen short of establish- tal muscle in both obese ob/ob and DIO mice. On the other ing this causal relationship. In obese New Zealand mice, hand, the LPD attenuated IR and BCAA/BCKA abundances reducing protein intake significantly improved glucose while probably reducing BCAA catabolism through lower- homeostasis, whereas BCAA repletion did not reverse it ing BCAA load to the catabolic pathway in ob/ob mice. (52). In a mouse model of preexisting DIO, reducing Furthermore, increasing BCAA intake promoted IR while dietary BCAAs improved glucoregulatory control while elevating BCAA/BCKA abundances and probably the cat- reducing body weight (53). In this study, however, the abolic flux. Therefore, IR is consistently attenuated by causal role of BCAAs remains unclear, because its impact reducing BCAA/BCKA abundances, rather than altering on glucose regulation could be attributed to changes in the status of catabolic flux, in obese mice. One recent body weight rather than BCAAs. In fatty Zucker rats, BCAA independent study reported similar beneficial effects of restriction showed no impact on systemic glucose toler- BT2 on insulin sensitivity in genetically obese rats (54). ance but improved skeletal muscle insulin sensitivity (7). Other reports have suggested that reducing BCAA intake Similarly, we found that a low-BCAA diet did not affect improves insulin sensitivity via various mechanisms in- systemic insulin sensitivity in ob/ob mice (Supplementary cluding reducing body weight or increasing energy expen- Fig. 11). However, reducing protein intake significantly diture (7,52,53). The common feature of these studies is improved systemic glucose tolerance in ob/ob mice. Re- the reduced abundances of BCAAs (and probably of ducing protein intake reduced the uptake of all amino acids BCKAs), corroborating the pathogenic roles of elevated and the plasma levels of eight amino acids, including BCAAs and BCKAs in obesity-associated IR. BCAAs, phenylalanine, tryptophan, glutamine, tyrosine, BCAA-stimulated mTORC1 activation has been widely and proline (Supplementary Fig. 10). Increasing BCAA implicated in reducing insulin signaling (2,7,56–58). Our intake in mice fed the LPD restored plasma BCAA levels data show that BCKAs are also capable of suppressing and significantly impaired systemic insulin sensitivity, insulin signaling. Thus, in addition to BCAAs per se, other specifying the causal role of BCAAs. The impacts of components of the BCAA catabolic pathway may also play non-BCAA amino acids affected by the LPD remain to important roles in obesity-associated IR. Furthermore, our be tested, whereas phenylalanine and tyrosine have been data suggest that mTORC1 probably mediates the sup- associated with IR (1). BCAA supplementation did not pression of AKT phosphorylation by BCAAs and BCKAs. reduce glucose tolerance to the level of that in the NPD However, disconnections have been observed between group, suggesting that the systemic insulin-sensitizing enhanced mTORC1 and impaired insulin signaling. For effect of BCAA restriction in obese ob/ob mice may need example, BCAA supplementation impaired insulin signal- assistance from other amino acids. Importantly, BCAA ing without enhancing mTORC1 activity in liver in ob/ob supplementation did not change the body weight of mice fed the LPD. The tissue-specific impacts of manipu- ob/ob mice fed the LPD. Compared with the total energy lations of insulin signaling and mTORC1 activity by BCAA intake (;94% from carbohydrate and fat in the LPD), the also remain puzzling. A recent study showed that BCAAs caloric intake from BCAAs is likely very small. Thus, the becoming replete while dietary protein is reduced restores caloric contribution from BCAA oxidation is unlikely to mTORC1 signaling but does not induce IR in obese New account for IR pathogenesis. Together with results show- Zealand mice (52). Thus, mTORC1 activation cannot fully ing that BT2 attenuated IR while enhancing BCAA oxida- explain the effects of modulating BCAA catabolism on tion without affecting food intake and body weight in both insulin sensitivity, and additional mTORC1-independent obese ob/ob and DIO mice (Figs. 4 and 7), these data clearly mechanisms warrant further investigation. demonstrate the causal role of a BCAA catabolic defect in The roles of BCAA catabolic flux and the distal steps of obesity-associated IR. On the other hand, although the BCAA catabolism in obesity-associated IR remain to be effects of BT2 on ob/ob mice are similar to those recently further investigated. In addition to the “clogging” model, reported in a genetic obese rat model (54) (Fig. 4), the recent studies have indicated that elevated BCAA oxidation beneficial impacts of BT2 in DIO mice (Fig. 7) add to the in skeletal muscle in obese db/db mice may lead to the clinical relevance of our findings, as the DIO rodents production of 3-hydroxyisobutyrate from valine and pro- remain the most clinically relevant model by mimicking mote IR (55,57). In contrast, another report suggested that the common cause and metabolic impairments of obesity defective BCAA metabolism in muscle contributes to im- in humans. paired lipid metabolism and thus the development of It has been suggested that, because of reduced catab- obesity-associated IR (59). BT2 treatment enhances olism by adipose tissue in obese mice, BCAA catabolism is BCAA catabolism in muscle in healthy mice (55), obese enhanced in liver and skeletal muscle, which “clogs” lipid rats (54), and obese mice (Figs. 4 and 7), while attenuating and glucose oxidation and promotes IR (50). A more recent IR in obese animals, supporting the detrimental effects of study showed that obese and insulin-resistant mice shunt impaired BCAA oxidation in muscle of obese animals. 1744 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

Interestingly, BT2 treatment increased BCKD activity inhibitor dichloroacetate (Supplementary Fig. 12). The (Figs. 4A and 7A) without reducing its phosphorylation structural constraints that preclude BT2 binding to in skeletal muscle of obese ob/ob mice (Supplementary Fig. PDKs are reflected by the absence of PDK inhibition and PDH 6) and obese rats (54). But BT2 increases the expression of activation by BT2 (data not shown). Therefore, the metabolic BCKD subunits (Supplementary Fig. 6) and multiple distal phenotypes resulting from BT2 treatment are unlikely to genes in the BCAA catabolic pathway (Supplementary Fig. be related to PDK-regulated PDH activity; rather, they are 7), which may contribute to enhanced BCAA catabolic flux. exerted via the reprogramming of BCAA catabolism. Quantitative measurements of BCAA catabolic flux at In summary, this study provides a compelling mecha- various steps in tissues of obese animals upon BT2 treat- nistic basis for the strong association between elevated ment or dietary manipulation will shed light on the roles plasma BCAAs and the onset of IR in humans, demonstrat- of BCAA catabolic flux and the distal steps in obesity- ing the causal role of a BCAA catabolic defect and the associated IR. pathogenic role of elevated BCAAs/BCKAs in obesity- Tissue-specificpatternsofBCAAcatabolismand associated IR. The influences of BCAA intake on insulin responses to BT2 treatment have been observed in various sensitivity indicate the potential impact of dietary protein experimental settings. In the bioinformatics analyses, the on IR and diabetes in obese people, consistent with a recent BCAA catabolic pathway in skeletal muscle showed a much report showing that protein restriction improves glycemic weaker association with IR than that in adipose tissue and control in overweight and mildly obese humans (60). liver. BCAA catabolic gene expression and phosphorylation BCKDK and its inhibitor represent new pharmacological patterns vary dramatically in various tissues of ob/ob mice. targets and an approach to attenuating obesity-associated BCAA abundances are elevated in adipose tissue, un- IR. Finally, the genetic link between IR and the BCAA changed in skeletal muscle, and reduced in liver in ob/ob catabolic pathway in human populations highlights the mice. Furthermore, BT2 treatment and dietary approaches clinical significance of BCAA catabolism as a potential ther- alter insulin signaling differently in liver, skeletal muscle, apeutic target in treating obesity-associated IR and diabetes. and adipose tissue. Interestingly, BT2 treatment affects BCKD activity, BCKD phosphorylation, BCKD subunits, and distal gene expression in a tissue-specific pattern. To Funding. This work was supported by the Ministry of Science and Technology make it even more complicated, one recent study showed of China (grant nos. 2012BAI02B05 and 2013YQ030923), the National Interna- that acute BT2 treatment in healthy mice activates BCAA tional Science Cooperation Foundation (grant no. 2015DFA30560), the National oxidation specifically in muscle even though BCKD phos- Natural Science Foundation of China (grant nos. NSFC81570717 and 81522011), phorylation is significantly reduced in liver, skeletal mus- the National Institutes of Health (grant nos. HL108186, HL103205, HL098954, and cle, and heart (55). These intriguing patterns of BCAA DK62306), the National Heart, Lung, and Blood Institute (grant no. HL080111), the catabolism and the impacts of BT2 may be attributed to Laubisch Fund (to the University of California Los Angeles), the Welch Foundation fi (grant no. I-1286), and the Science and Technology Commission of Shanghai the tissue-speci c expression of BCAA catabolic enzymes Municipality (grant nos. 13ZR1423300 and 16JC1404400). L.S. is supported by and the pharmacokinetics of BT2. For example, although a China Scholarship Council scholarship, a UCLA Eureka and Hyde scholarship, BCKDK expression is high in liver, skeletal muscle, and and a Burroughs Wellcome Fund fellowship. X.Y. is supported by the National adipose tissue, BCKD phosphatase (PP2Cm) expression is Institutes of Health/National Institute of Diabetes and Digestive and Kidney lower in skeletal muscle and adipose tissue than in the Diseases (grant no. R01DK104363), the American Heart Association (grant no. liver. The strong dephosphorylation of BCKD by BT2 in 13SDG17290032), an American Heart Association Cardiovascular Genome- liver might be the result of both BCKDK inhibition and Phenome Study Pathway Grant, and a Leducq Foundation Transatlantic Networks PP2Cm activation. Nevertheless, despite these tissue- of Excellence Grant. specific patterns of BCAA catabolism and BT2 effects, Duality of Interest. Y.W. and H.S. participated in an advisory board for fl systemic BCAA catabolism is defective in obese mice but Ramino Bio Ltd. No other potential con icts of interest relevant to this article were can be relieved by BT2 treatment, irrespective of the site, reported. Author Contributions. M.Z., J.S., C.-Y.W., L.S., W.D., Y.L., M.C., R.M.W., in order to improve systemic insulin sensitivity. Mean- Jiq.W., Ji W., and W.-J.G. performed the research. M.Z., J.S., C.-Y.W., L.S., W.D., while, more mechanistic details remain to be determined. Y.L., M.C., R.M.W., Jiq.W., Ji W., W.-J.G., X.Q., A.J.L., Z.L., W.W., G.N., X.Y., BT2 is a potent allosteric inhibitor of BCKDK. BCKDK D.T.C., Y.W., and H.S. prepared the manuscript. M.Z., C.-Y.W., L.S., W.D., Y.W., and (PDKs) share similar and H.S. analyzed the data. X.Q. performed the chemical design and synthesis. structures. PDKs phosphorylate and inhibit the pyruvate A.J.L., Z.L., W.W., G.N., X.Y., D.T.C., and Y.W. designed the overall study and dehydrogenase (PDH) complex, a key player in glucose analyzed the data. H.S. designed the research. H.S. is the guarantor of this work metabolism. However, the allosteric ligand-binding pocket and, as such, had full access to all the data in the study and takes responsibility in BCKDK (volume 412Å3) is two- to fourfold larger than for the integrity of the data and the accuracy of the data analysis. 3– 3 its counterparts in PDKs (90Å 213Å ) (Supplementary References Fig. 12). The larger allosteric ligand-binding volume in 1. Newgard CB, An J, Bain JR, et al. A branched-chain amino acid-related BCKDK allows for the occupancy of natural inhibitors like metabolic signature that differentiates obese and lean humans and contributes to BCKAs or the artificial inhibitor BT2. By comparison, the insulin resistance. Cell Metab 2009;9:311–326 allosteric sites in PDKs can accommodate up to 3-carbon 2. Lynch CJ, Adams SH. Branched-chain amino acids in metabolic signalling a-keto acids such as pyruvate and its small-molecule and insulin resistance. Nat Rev Endocrinol 2014;10:723–736 diabetes.diabetesjournals.org Zhou and Associates 1745

3. Felig P, Marliss E, Cahill GFJ Jr. Plasma amino acid levels and insulin Obesity: implications for dietary guidelines (NUGENOB) randomized trial. Am J Clin secretion in obesity. N Engl J Med 1969;281:811–816 Nutr 2017;106:902–908 4. Lu J, Xie G, Jia W, Jia W. Insulin resistance and the metabolism of branched- 26. Dupuis J, Langenberg C, Prokopenko I, et al.; DIAGRAM Consortium; GIANT chain amino acids. Front Med 2013;7:53–59 Consortium; Global BPgen Consortium; Anders Hamsten on behalf of Procardis 5. Lu Y, Wang Y, Ong C-N, et al. Metabolic signatures and risk of type 2 diabetes Consortium; MAGIC investigators. New genetic loci implicated in fasting glucose in a Chinese population: an untargeted metabolomics study using both LC-MS and homeostasis and their impact on type 2 diabetes risk. Nat Genet 2010;42:105–116 GC-MS. Diabetologia 2016;59:2349–2359 27. Knowles JW, Xie W, Zhang Z, et al.; RISC (Relationship between Insulin 6. Chen T, Ni Y, Ma X, et al. Branched-chain and aromatic amino acid profiles Sensitivity and Cardiovascular Disease) Consortium; EUGENE2 (European and diabetes risk in Chinese populations. Sci Rep 2016;6:20594 Network on Functional Genomics of Type 2 Diabetes) Study; GUARDIAN 7. White PJ, Lapworth AL, An J, et al. Branched-chain amino acid restriction in (Genetics UndeRlying DIAbetes in HispaNics) Consortium; SAPPHIRe (Stanford Zucker-fatty rats improves muscle insulin sensitivity by enhancing efficiency of Asian and Pacific Program for Hypertension and Insulin Resistance) Study. fatty acid oxidation and acyl-glycine export. Mol Metab 2016;5:538–551 Identification and validation of N-acetyltransferase 2 as an insulin sensitivity 8. She P, Olson KC, Kadota Y, et al. Leucine and protein metabolism in obese gene [published correction appears in J Clin Invest 2016;126:403]. J Clin Invest Zucker rats. PLoS One 2013;8:e59443 2015;125:1739–1751 9. Menni C, Fauman E, Erte I, et al. Biomarkers for type 2 diabetes and impaired 28. Mäkinen VP, Civelek M, Meng Q, et al.; Coronary Artery Disease Genome- fasting glucose using a nontargeted metabolomics approach. Diabetes 2013;62: Wide Replication and Meta-Analysis (CARDIoGRAM) Consortium. Integrative ge- 4270–4276 nomics reveals novel molecular pathways and gene networks for coronary artery 10. Wang TJ, Larson MG, Vasan RS, et al. Metabolite profiles and the risk of disease. PLoS Genet 2014;10:e1004502 developing diabetes. Nat Med 2011;17:448–453 29. International HapMap Consortium. The International HapMap Project. Nature 11. Yu D, Moore SC, Matthews CE, et al. Plasma metabolomic profiles in as- 2003;426:789–796 sociation with type 2 diabetes risk and prevalence in Chinese adults. Metabolomics 30. 1000 Genomes Project Consortium; Abecasis GR, Auton A, Brooks LD, et al. 2016;12. pii: 3 An integrated map of genetic variation from 1,092 human genomes. Nature 2012; 12. Würtz P, Soininen P, Kangas AJ, et al. Branched-chain and aromatic amino acids 491:56–65 are predictors of insulin resistance in young adults. Diabetes Care 2013;36:648–655 31. Arneson D, Bhattacharya A, Shu L, Mäkinen VP, Yang X. Mergeomics: a web 13. Floegel A, Stefan N, Yu Z, et al. Identification of serum metabolites associated server for identifying pathological pathways, networks, and key regulators via with risk of type 2 diabetes using a targeted metabolomic approach. Diabetes multidimensional data integration. BMC Genomics 2016;17:722 2013;62:639–648 32. Langfelder P, Horvath S. WGCNA: an R package for weighted correlation 14. Melnik BC. Leucine signaling in the pathogenesis of type 2 diabetes and network analysis. BMC Bioinformatics 2008;9:559 obesity. World J Diabetes 2012;3:38–53 33. Kanehisa M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes. 15. Shah SH, Crosslin DR, Haynes CS, et al. Branched-chain amino acid levels Nucleic Acids Res 2000;28:27–30 are associated with improvement in insulin resistance with weight loss. Dia- 34. Shu L, Zhao Y, Kurt Z, et al. Mergeomics: multidimensional data integration to betologia 2012;55:321–330 identify pathogenic perturbations to biological systems. BMC Genomics 2016;17: 16. Laferrère B, Reilly D, Arias S, et al. Differential metabolic impact of gastric 874 bypass surgery versus dietary intervention in obese diabetic subjects despite 35. Zhu J, Wiener MC, Zhang C, et al. Increasing the power to detect causal identical weight loss. Sci Transl Med 2011;3:80re2 associations by combining genotypic and expression data in segregating pop- 17. Hutson SM, Sweatt AJ, Lanoue KF. Branched-chain [corrected] amino acid ulations. PLoS Comput Biol 2007;3:e69 metabolism: implications for establishing safe intakes. J Nutr 2005;135(Suppl.): 36. Zhu J, Zhang B, Smith EN, et al. Integrating large-scale functional genomic 1557S–1564S data to dissect the complexity of yeast regulatory networks. Nat Genet 2008;40: 18. Lu G, Sun H, She P, et al. Protein phosphatase 2Cm is a critical regulator of 854–861 branched-chain amino acid catabolism in mice and cultured cells. J Clin Invest 37. Bennett BJ, Farber CR, Orozco L, et al. A high-resolution association mapping 2009;119:1678–1687 panel for the dissection of complex traits in mice. Genome Res 2010;20:281–290 19. Burrage LC, Nagamani SC, Campeau PM, Lee BH. Branched-chain amino 38. Parks BW, Nam E, Org E, et al. Genetic control of obesity and gut microbiota acid metabolism: from rare Mendelian diseases to more common disorders. Hum composition in response to high-fat, high-sucrose diet in mice. Cell Metab 2013; Mol Genet 2014;23:R1–R8 17:141–152 20. Oyarzabal A, Martínez-Pardo M, Merinero B, et al. A novel regulatory defect in 39. Sun H, Olson KC, Gao C, et al. Catabolic defect of branched-chain amino the branched-chain a-keto acid dehydrogenase complex due to a mutation in the acids promotes heart failure. Circulation 2016;133:2038–2049 PPM1K gene causes a mild variant phenotype of maple syrup urine disease. Hum 40. Tso S-C, Gui W-J, Wu C-Y, et al. Benzothiophene carboxylate derivatives as Mutat 2013;34:355–362 novel allosteric inhibitors of branched-chain a-ketoacid dehydrogenase kinase. J 21. Zinnanti WJ, Lazovic J, Griffin K, et al. Dual mechanism of brain injury and Biol Chem 2014;289:20583–20593 novel treatment strategy in maple syrup urine disease. Brain 2009;132:903–918 41. Yang X, Zhang B, Molony C, et al. Systematic genetic and genomic analysis of 22. Tiffin N, Adie E, Turner F, et al. Computational disease gene identification: P450 enzyme activities in human liver. Genome Res 2010;20:1020– a concert of methods prioritizes type 2 diabetes and obesity candidate genes. 1036 Nucleic Acids Res 2006;34:3067–3081 42. Wang IM, Zhang B, Yang X, et al. Systems analysis of eleven rodent disease 23. Taneera J, Lang S, Sharma A, et al. A systems genetics approach identifies genes models reveals an inflammatome signature and key drivers. Mol Syst Biol 2012;8: and pathways for type 2 diabetes in human islets. Cell Metab 2012;16:122–134 594 24. Xu M, Qi Q, Liang J, et al. Genetic determinant for amino acid metabolites and 43. Meng Q, Ying Z, Noble E, et al. Systems nutrigenomics reveals brain gene changes in body weight and insulin resistance in response to weight-loss diets: the networks linking metabolic and brain disorders. EBioMedicine 2016;7:157–166 Preventing Overweight Using Novel Dietary Strategies (POUNDS LOST) trial. 44. Tai ES, Tan MLS, Stevens RD, et al. Insulin resistance is associated with Circulation 2013;127:1283–1289 a metabolic profile of altered protein metabolism in Chinese and Asian-Indian men. 25. Goni L, Qi L, Cuervo M, et al. Effect of the interaction between diet com- Diabetologia 2010;53:757–767 position and the PPM1K genetic variant on insulin resistance and b cell function 45. She P, Van Horn C, Reid T, Hutson SM, Cooney RN, Lynch CJ. Obesity-related markers during weight loss: results from the Nutrient Gene Interactions in Human elevations in plasma leucine are associated with alterations in enzymes involved in 1746 Targeting BCAA Catabolism to Treat IR Diabetes Volume 68, September 2019

branched-chain amino acid metabolism. Am J Physiol Endocrinol Metab 2007; 53. Cummings NE, Williams EM, Kasza I, et al. Restoration of metabolic health by 293:E1552–E1563 decreased consumption of branched-chain amino acids. J Physiol 2018;596:623– 46. Pietiläinen KH, Naukkarinen J, Rissanen A, et al. Global transcript profiles of 645 fat in monozygotic twins discordant for BMI: pathways behind acquired obesity. 54. White PJ, McGarrah RW, Grimsrud PA, et al. The BCKDH kinase and PLoS Med 2008;5:e51 phosphatase integrate BCAA and lipid metabolism via regulation of ATP-citrate 47. Lackey DE, Lynch CJ, Olson KC, et al. Regulation of adipose branched-chain . Cell Metab 2018;27:1281–1293.e7 amino acid catabolism enzyme expression and cross-adipose amino acid flux in 55. Neinast MD, Jang C, Hui S, et al. Quantitative analysis of the whole-body human obesity. Am J Physiol Endocrinol Metab 2013;304:E1175–E1187 metabolic fate of branched-chain amino acids. Cell Metab 2019;29:417– 48. Herman MA, She P, Peroni OD, Lynch CJ, Kahn BB. Adipose tissue branched 429.e4 chain amino acid (BCAA) metabolism modulates circulating BCAA levels. J Biol 56. Xiao F, Huang Z, Li H, et al. Leucine deprivation increases hepatic insulin Chem 2010;285:11348–11356 sensitivity via GCN2/mTOR/S6K1 and AMPK pathways. Diabetes 2011;60:746–756 49. Zimmerman HA, Olson KC, Chen G, Lynch CJ. Adipose transplant for inborn 57. Jang C, Oh SF, Wada S, et al. A branched-chain amino acid metabolite drives errors of branched chain amino acid metabolism in mice. Mol Genet Metab 2013; vascular fatty acid transport and causes insulin resistance. Nat Med 2016;22:421– 109:345–353 426 50. Newgard CB. Interplay between lipids and branched-chain amino acids in 58. Xiao F, Yu J, Guo Y, et al. Effects of individual branched-chain amino acids development of insulin resistance. Cell Metab 2012;15:606–614 deprivation on insulin sensitivity and glucose metabolism in mice. Metabolism 51. Bifari F, Nisoli E. Branched-chain amino acids differently modulate catabolic 2014;63:841–850 and anabolic states in mammals: a pharmacological point of view. Br J Pharmacol 59. Lerin C, Goldfine AB, Boes T, et al. Defects in muscle branched-chain amino 2017;174:1366–1377 acid oxidation contribute to impaired lipid metabolism. Mol Metab 2016;5:926– 52. Maida A, Chan JSK, Sjøberg KA, et al. Repletion of branched chain amino 936 acids reverses mTORC1 signaling but not improved metabolism during dietary 60. Fontana L, Cummings NE, Arriola Apelo SI, et al. Decreased consumption of protein dilution. Mol Metab 2017;6:873–881 branched-chain amino acids improves metabolic health. Cell Rep 2016;16:520–530